More Like this

1. Accelerating Simulation of Particle Trajectories in Microfluidic Devices by Constructing a Cloud Database

   https://doi.org/10.1109/ISVLSI.2018.00126  

   Wang, Junchao; Fu, Lingxuan; Yu, Liyang; Huang, Xiwei; Brisk, Philip; Grover, William H. (July 2018, 2018 IEEE Computer Society Annual Symposium on VLSI (ISVLSI))

   Microfluidic cell sorters have shown great potential to revolutionize the current technique of enriching rare cells. In the past decades, different microfluidic cell sorters have been developed by researchers for separating circulating tumor cells, T-cells, and other biological markers from blood samples. However, it typically takes months or even years to design these microfluidic cell sorters by hand. Thus, researchers tend to use computer simulation (usually finite element analysis) to verify their designs before fabrication and experimental testing. Despite this, conducting precision finite element analysis of microfluidic devices is computationally expensive and labor-intensive. To address this issue, we recently presented a microfluidic simulation method that can simulate the behavior of fluids and particles in some typical microfluidic chips instantaneously. Our method decomposes the chip into channels and intersections. The behavior of fluid in each channel is determined by leveraging analogies with electronic circuits, and the behavior of fluid and particles in each intersection is determined by querying a database containing 92,934 pre-simulated channel intersections. While this approach successfully predicts the behavior of complex microfluidic chips in a fraction of the time required by existing techniques, we nonetheless identified three major limitations with this method: (1) the library of pre-simulated channel intersections is unnecessarily large (only 2,072 of 92,934 were used); (2) the library contains only cross-shaped intersections (and no other intersection geometries); and (3) the range of fluid flow rates in the library is limited to 0 to 2 cm/s. To address these deficiencies, in this work we present an improved method for instantaneously simulating the trajectories of particles in microfluidic chips. Firstly, inspired by dynamic programming, our new method optimizes the generation of pre-simulated intersection units and avoids generating unnecessary simulations. Secondly, we constructed a cloud database (http://cloud.microfluidics.cc) to share our pre-simulated results and to let users become contributors and upload their simulation results into the cloud database as a benefit to the whole microfluidic simulation community. Lastly, we investigated the impact of different channel angles and different fluid flow rates on predicting the trajectories of particles. We found a wide range of device geometries and flow rates over which our existing simulation results can be extended without having to perform additional simulations. Our method should accelerate the simulation of particles in microfluidic chips and enable researchers to design new microfluidic cell sorter chips more efficiently. 

   more » « less
2. Complementary performances of convolutional and capsule neural networks on classifying microfluidic images of dividing yeast cells

   https://doi.org/10.1371/journal.pone.0246988  

   Ghafari, Mehran; Clark, Justin; Guo, Hao-Bo; Yu, Ruofan; Sun, Yu; Dang, Weiwei; Qin, Hong (March 2021, PLOS ONE)

   Schwenker, Friedhelm (Ed.)

   Microfluidic-based assays have become effective high-throughput approaches to examining replicative aging of budding yeast cells. Deep learning may offer an efficient way to analyze a large number of images collected from microfluidic experiments. Here, we compare three deep learning architectures to classify microfluidic time-lapse images of dividing yeast cells into categories that represent different stages in the yeast replicative aging process. We found that convolutional neural networks outperformed capsule networks in terms of accuracy, precision, and recall. The capsule networks had the most robust performance in detecting one specific category of cell images. An ensemble of three best-fitted single-architecture models achieves the highest overall accuracy, precision, and recall due to complementary performances. In addition, extending classification classes and data augmentation of the training dataset can improve the predictions of the biological categories in our study. This work lays a useful framework for sophisticated deep-learning processing of microfluidic-based assays of yeast replicative aging. 

   more » « less
3. Self-controlling photonic-on-chip networks with deep reinforcement learning

   https://doi.org/10.1038/s41598-021-02583-7  

   Do, Nguyen; Truong, Dung; Nguyen, Duy; Hoai, Minh; Pham, Cuong (December 2021, Scientific Reports)

   Abstract We present a novel photonic chip design for high bandwidth four-degree optical switches that support high-dimensional switching mechanisms with low insertion loss and low crosstalk in a low power consumption level and a short switching time. Such four-degree photonic chips can be used to build an integrated full-grid Photonic-on-Chip Network (PCN). With four distinct input/output directions, the proposed photonic chips are superior compared to the current bidirectional photonic switches, where a conventionally sizable PCN can only be constructed as a linear chain of bidirectional chips. Our four-directional photonic chips are more flexible and scalable for the design of modern optical switches, enabling the construction of multi-dimensional photonic chip networks that are widely applied for intra-chip communication networks and photonic data centers. More noticeably, our photonic networks can be self-controlling with our proposed Multi-Sample Discovery model, a deep reinforcement learning model based on Proximal Policy Optimization. On a PCN, we can optimize many criteria such as transmission loss, power consumption, and routing time, while preserving performance and scaling up the network with dynamic changes. Experiments on simulated data demonstrate the effectiveness and scalability of the proposed architectural design and optimization algorithm. Perceivable insights make the constructed architecture become the self-controlling photonic-on-chip networks. 

   more » « less
4. Decoding of Code-Multiplexed Coulter Sensor Signals via Deep Learning

   https://doi.org/10.1109/TRANSDUCERS.2019.8808195  

   Wang, Ningquan; Liu, Ruxiu; Asmare, Norh; Anandakumar, Dakshitha B; Sarioglu, A. Fatih (June 2019, Transducers 2019)

   Code-multiplexed Coulter sensors can easily be integrated into microfluidic devices and provide information on spatiotemporal manipulations of suspended particles for quantitative sample assessment. In this paper, we introduced a deep learning-based decoding algorithm to process the output waveform from a network of code- multiplexed Coulter sensors on a microfluidic device. Our deep learning-based algorithm both simplifies the design of coded Coulter sensors and increases the signal processing speed. As a proof of principle, we designed and fabricated a microfluidic platform with 10 code-multiplexed Coulter sensors, and used a suspension of human ovarian cancer cells as a test sample to characterize the system. Our deep learning-based algorithm resulted in an 87% decoding accuracy at a sample processing speed of 800 particles/s. 

   more » « less
5. Common Food-Wrap Film as a Cost-Effective and Readily Available Alternative to Thermoplastic Polyurethane (TPU) Membranes for Microfluidics On-Chip Valves and Pumps

   https://doi.org/10.31224/4128  

   Nguyen, Huu_Anh Minh; Volosov, Mark; Maffei, Jessica; Voronov, Roman (November 2024, engrxiv)

   Recently, there has been an increasing effort in developing new fabrication methods for rapid prototyping of microfluidic chips using laser cutting and 3D printing. However, although these approaches can readily generate rigid parts of the devices, it is not trivial to integrate flexible components (e.g. on-chip valve and/or pump membranes) within the same build. This has led to the recent adoption of thermoplastic polyurethane (TPU) membranes sandwiched between the rigid layers to introduce the necessary flexibility to the chips. Despite its utility, TPU is not without its challenges—it is relatively expensive and surprisingly difficult to source. To overcome these difficulties, our study introduces the use of common food wrapping film as a cost-effective and readily available alternative to TPU, demonstrating its compatibility in fabricating essential microfluidic components such as on-chip valves and peristaltic pumps. Our findings show that this alternative maintains the performance standards required for sophisticated microfluidic applications while significantly alleviating logistical and financial constraints. The results show high cyclability of the membrane, up to 850,000 in continuous testing conditions, at 1 Hz, while also can block the fluid flow at as low as 250 kPa. Regarding the micropumps, it was shown that adequate flow rate of around 4 μL/min can be achieved. 

   more » « less